How each check works,and where it stops.

Contract Risk Scanner

1. eth_getCode at the latest block. The code is walked opcode by opcode, skipping PUSH data, after cutting recognised compiler metadata.
2. Every PUSH4 constant is a candidate selector. A risky function is reported when its selector is among them (selectors computed from the signatures below).
3. Proxy slots read with eth_getStorageAt: implementation 0x360894a13b…, admin 0xb53127684a…, beacon 0xa3f0ad74e5…, EIP-1822 0xc5f16f0fcc…; EIP-1167 minimal proxies and EIP-7702 delegations by their code pattern. A proxy's implementation code is scanned too.
4. owner(), getOwner(), paused(), name(), symbol(), decimals(), totalSupply() read with eth_call; the owner's code tells an account from a contract.
5. Source, contract name and creator from the Elysium explorer API; verified source lines matching each power are quoted with file and line.
6. Reports on the address from the report ledger.

• This scan reads code and storage. It does not simulate transactions and does not run the contract.
• A function found in the code may be guarded, unused or harmless; a function not found may still exist under another name.
• A proxy can be upgraded after this scan. Scan again before every large interaction.
• Roles held through AccessControl, multisigs or timelocks are not resolved to people.
• No flag means nothing was found by these checks. It does not mean the contract is safe.

Wallet Permission Inspector

Approval (ERC-20 with three topics, ERC-721 with four) and ApprovalForAll logs with the wallet as owner, from block 0 to the head. A range answered with "too many logs" is split in two, newest half first, up to 48 times; what still fails is listed as a gap. Then allowance(), isApprovedForAll() or getApproved() read now. Unlimited means 2^255 or more. Revoke sends approve(spender, 0), setApprovalForAll(operator, false) or approve(0x0, tokenId) from your wallet.

Threat Monitoring rules

1. highUnlimited approval granted (unlimited-approval): An outgoing approve() or increaseAllowance() whose amount is 2^255 or more.
2. highCollection operator granted (operator-approval): An outgoing setApprovalForAll(operator, true).
3. cautionLarge HYPE transfer out (large-out): One outgoing transaction carrying 100 HYPE or more.
4. infoNew counterparty (new-counterparty): An outgoing transaction in the last 24 h to an address this wallet never sent to earlier in the fetched history (needs at least 20 older transactions).
5. cautionBurst of transactions (burst): 10 or more outgoing transactions within 10 minutes.
6. cautionFailed transactions in a row (failed-streak): 3 or more consecutive failed outgoing transactions.
7. highMany tokens sent out quickly (token-sweep): 3 or more different tokens sent out within 30 minutes (a pattern of drainer scripts).
8. cautionCalled a reported contract (reported-contract): An outgoing transaction to a contract with an open or upheld report on the report ledger.
9. cautionLook-alike sender (address poisoning) (address-poisoning): An incoming token transfer of 0 from an address whose first 4 and last 4 hex characters match an address this wallet sent to, but which is a different address.

Defaults shown; each threshold can be changed on the monitor page. Data: the latest 100 transactions and 100 token transfers per wallet from the Elysium explorer API.

Contract events

• highUpgraded(address)
• highBeaconUpgraded(address)
• highAdminChanged(address,address)
• cautionOwnershipTransferred(address,address)
• highPaused(address)
• infoUnpaused(address)
• cautionRoleGranted(bytes32,address,address)
• infoRoleRevoked(bytes32,address,address)
• cautionTransfer(0x0, to, amount) (a mint)

Transaction Risk Preview

The calldata is decoded with a short standard list (approve, increaseAllowance, transfer, transferFrom, permit, setApprovalForAll, safeTransferFrom, multicall and a few more), else with the target's verified ABI from the explorer, or its implementation's. Multicall inner calls are decoded one level deep. The spender's code is read for approvals. The target gets a full scan. The preview reads the request and the target. It does not simulate the transaction.

Report ledger

SecurityumReportLedger, live on Elysium at 0xfdd15018e685996089f9ae596e8a7cb104aaf7d6. Bond, resolver key and the two windows are fixed at deployment and readable on chain: bond 0.05 HYPE per report and per counter, resolver key 0x98a5404a8224c0e43aa622c53b9e3059d24a3ea8, counter-evidence window 3 days, then verdict window 4 days. One open report per reporter per contract; up to 8 counter posts per report; evidence is a 32-byte hash plus an optional link of up to 256 bytes. No owner, fee, pause, upgrade or sweep.

Planned, not built

• plannedHosted alerts. Alerts delivered while the site is closed (push, email or webhook). Today alerts are evaluated in your browser while a page is open.
• plannedDeveloper API. The scanner and rules as an API with keys and quotas. Today the read endpoints of this site are rate-limited and keyless.
• plannedSignature preview. Decoding of typed-data signature requests (permits, orders) before you sign them. Today the preview covers transactions only.
• planned$SECURITYUM for quotas. Using the token for API and hosted-alert quotas once those services exist. Planned and unproven; nothing needs the token today.